Stefano Ciapponi

Signal Processing Researcher & PhD Candidate specializing in algorithm-system co-design for edge computing systems. I develop end-to-end solutions for temporal signals in edge computing scenarios, with a focus on bioacoustics conservation technology.

“I’m a professional jack-of-all-trades who designs messy systems to analyze even messier systems.”

Current Roles

• Signal Processing Researcher @ Fondazione Bruno Kessler (FBK)
• PhD Candidate @ University of Trento
• Teaching Assistant @ University of Trento
  • Introduction to Computer Programming (C++)
  • Computer Architectures

Research & Projects

Conservation Technology

Enabling Multi-Species Bird Classification on Low-Power Bioacoustic Loggers (WrenNet)

Species-aware bird classifier for resource-constrained acoustic monitoring devices.

• Achievement: Efficient species classification on AudioMoth devices
• Key Features:
  • Fancy DDSP semi-learnable frequency warping.
  • 16x more efficient on RPi3 compared to Birdnet.
  • Works on Audiomoth!
• Paper
• Repository

Human-Centered Audio Processing

Exploiting NACs

Research on Neural Audio Codec optimization for edge to gateway compression.

• Achievement: Real-time audio compression on RPi 3B+ with downstream Intent Classification and ASR
• GitHub Organization
• Paper

TinyVOCOS

Neural vocoder optimized for MCU deployment, showcasing efficient speech synthesis at the edge.

• Achievement: Successfully runs on Arm Cortex M7 MCUs
• Technologies: Neural Networks, MCUs, Edge Computing
• Project Details →
• Paper

Expertise

• Core Skills: Signal Processing, Edge Computing, Neural Networks
• Programming: C++, Assembly Language, Python
• Systems: Embedded Systems, MCUs, Distributed Computing
• Languages: Italian (Native), English (Native/Bilingual)

Education

• PhD in Electrical and Computer Engineering, University of Trento (2023-2026)
• MS in Information and Computer Engineering, University of Bologna (2021-2023)
• BS in Computer Science, University of Milano-Bicocca (2018-2021)

Publications

• Enabling Multi-Species Bird Classification on Low-Power Bioacoustic Loggers (PREPRINT) (2025)
  • Conservation Technology
  • arXiv Preprint
• TinyVOCOS: Neural Vocoders on MCUs (2023)
  • Human-Centered Audio Processing
  • IEEE Publication
  • Project Details
• Exploiting Neural Audio Codecs (2025)
  • Human-Centered Audio Processing
  • Published at EUSIPCO 2025
  • Paper
  • View Paper